- Current status
Build and test results
- Nios II
- PowerPC (32-bit soft-float)
- PowerPC (32-bit hard-float)
- PowerPC (64-bit hard-float)
- PowerPC64LE (64-bit hard-float)
- RISC-V (rv64imac/lp64)
- RISC-V (rv64imafdc/lp64)
- RISC-V (rv64imafdc/lp64d)
- S/390 (32-bit)
- S/390 (64-bit)
- SPARC (32-bit)
- SPARC (64-bit)
- x86 (32-bit, Linux)
- x86_64 (64-bit, Linux)
- x86_64 (x32, Linux)
- x86 (32-bit, Hurd)
- Packaging Changes
1. Current status
The release branch of glibc-2.28 is currently in active development.
What things do we want to accomplish this release?
2.1. Release blockers?
- Intel CET enablement (H.J. to add patches to this entry for review).
regex equivalence class regression Fixed by 7b2f4cedf044ea83f53f6b43a5bf6871eb9ce969 (Florian Weimer).
Due to the the Unicode/CLDR data updates we have a regression in the equivalence class API.
- Malloc hook removal (DJ's work, Carlos reviewing).
2.2. Desirable this release?
Fix race in pthread_mutex_lock while promoting to PTHREAD_MUTEX_ELISION_NP BZ #23275 (Patch from Stefan, Carlos asked me to add it to the release blockers)
statx system call wrapper Fixed by fd70af45528d59a00eb3190ef6706cb299488fcd (Florian Weimer).
2.3. Deferred to the next release?
Deferring arc4random addition to 2.29 to allow for more review.
- IBM long double ABI transition.
Optimized math routines, improved generic math functions.
Add 64-bit time for all architectures (added to blockers with kind permission from Carlos, including updates requested by Paul)
3. Build and test results
Describe build and test issues for each architecture, or confirm a clean build with no testsuite failures. The list below is not a complete list of ABI variants; testing should try to cover the different ABI variants as far as possible.
Build system: UNAME -a, GCC?, Binutils?, Kernel ?
TRIMMED LIST OF FAILURES.
On systems with a broken libidn2 (version < 2.0.5) the following tests fail:
FAIL: resolv/tst-resolv-ai_idn FAIL: resolv/tst-resolv-ai_idn-latin1
Linux between 4.13 and 4.15 return EOVERFLOW for LFS OFD locks usage in compat mode (non-LFS ABI running on a LFS default kernel, such as i386 on a x86_64 kernel or s390-32 on a s390-64 kernel). It was discussed in a libc-alpha thread. This is a kernel issue because NR_fcntl64 is the expected way to use OFD locks and it is used on GLIBC for both fcntl and fcntl64. So, the following tests fails when running in the mentioned kernel versions:
3.10. Nios II
3.11. PowerPC (32-bit soft-float)
3.12. PowerPC (32-bit hard-float)
3.13. PowerPC (64-bit hard-float)
3.14. PowerPC64LE (64-bit hard-float)
3.15. RISC-V (rv64imac/lp64)
3.16. RISC-V (rv64imafdc/lp64)
3.17. RISC-V (rv64imafdc/lp64d)
3.18. S/390 (32-bit)
3.19. S/390 (64-bit)
3.21. SPARC (32-bit)
3.22. SPARC (64-bit)
3.23. x86 (32-bit, Linux)
3.24. x86_64 (64-bit, Linux)
3.25. x86_64 (x32, Linux)
3.26. x86 (32-bit, Hurd)
4. Packaging Changes
4.1. The locale-archive file is much bigger
The iso14651_t1_common file, which contains the default collation on top of which the individual locales add their locale specific rules, has been updated to a new upstream version from ISO which is in sync with Unicode 9.0.0. The version of that file which was in glibc until now was very old and a huge amount of characters has been added to Unicode in the meantime. Therefore, this iso14651_t1_common file became much larger. As a consequence, the LC_COLLATE part of all locales became much larger. As LC_COLLATE makes up the bulk of the locale data, the size of the locales increased a lot. The locale-archive file which contains the data for all localess grew from 126 MiB to 206 Mib.
4.2. libidn2 is a recommended dependency
The getaddrinfo function, when called with the AI_IDN or AI_CANONIDN flags, will use the system libidn2 library to perform IDNA encoding. Version 2.0.5 or later is recommended, otherwise there will be some failures in the glibc test suite.
4.3. Change 3
And another change.